ORIGINAL ARTICLE

Evaluation of Next Generation Sequencing for Detecting HER2 CopyNumber in Breast and Gastric Cancers

Dongfeng Niu1 & Lei Li2 & Yang Yu2 & Wanchun Zang2 & Zhongwu Li1 & Lixin Zhou1 & Ling Jia1 & Guanhua Rao2 &Lianju Gao2 & Gang Cheng2 & Ke Ji3 & Dongmei Lin1

Received: 25 May 2020 /Accepted: 10 June 2020# The Author(s) 2020

AbstractAmplicon-based next generation sequencing (NGS) approaches have been preferentially adopted by the clinical laboratories onthe basis of a short turnaround time (TAT) and small DNA input needs. However, little work has been done to assess theamplicon-based NGS methods for copy number variation (CNV) detection in comparison with current standard methods likeimmunohistochemistry (IHC) and fluorescence in situ hybridization (FISH). The correlation between NGS based CNV detectionand the later standard methods has remained unexplored. We developed an amplicon-based panel to detect human epidermalreceptor growth factor (HER2) amplification in formalin-fixed paraffin-embedded (FFPE) tumor tissue samples from 280 breastcancer and 50 gastric cancer patients. Assessment by IHC and FISH was conducted in parallel, and descriptive statistics wereused to assess the concordance. The copy number detected by NGS was correlated with either the average HER2 copy number(signals/cell) (r = 0.844; p < 0.001) or the HER2/CEP17 ratio (r = 0.815; p < 0.001). We determined a cut-off value for NGS tocategorize HER2 amplification status by using 151 HER2 non-amplified FFPE samples. In breast cancer patients, the cut-offvalue was 2.910, with 95.35%, 98.67% and 97.29% sensitivity, specificity and concordance, respectively. However, this cut-offvalue displayed low sensitivity in gastric cancer patients (64.71%), and the following macrodissection procedure was noteffective for increasing sensitivity (57.14%). Evaluation of HER2 copy number with NGS in our study was comparable withIHC and FISH in breast cancer patients, but concordance in gastric cancer was only moderate. The greater discordance in gastriccancer may reflect the underlying biological mechanisms, and further study is warranted. NGS-based HER2 assessment maydecrease the equivocal HER2 determinations in breast cancer patients assessed by FISH/IHC.

Keywords Next generation sequencing .HER2 amplification . Breast cancer . Gastric cancer . FISH/IHC

Background

The reliable identification of clinically actionable genomicalteration method is critical for the precision cancer thera-py guidance. Next generation sequencing (NGS) has thecapability to simultaneously assess multiple genes with alimited biopsy material, thus representing both a cost andtissue-efficient alternative to current single-gene assess-ment methods [1–3]. Prior studies have shown that NGSenables a reliable detection for copy number variations(CNV) from the same assays used to detect sequence alter-ations, but less information is available on amplicon-basedtarget sequences [4–7]. CNV calling in the amplicon se-quence relies on the calculation of amplicon coverage andsuitable normalization. Several factors influence CNV de-tection, including the number of amplicons per gene, aver-age read dept., and tumor purity within the sample [7].

Dongfeng Niu, Lei Li and Yang Yu contributed equally to this work.

Electronic supplementary material The online version of this article(https://doi.org/10.1007/s12253-020-00844-w) contains supplementarymaterial, which is available to authorized users.

* Ke Jiji_jiafu@hotmail.com

* Dongmei LinDongm_lin@126.com

1 Department of Pathology, Key laboratory of Carcinogenesis andTranslational Research (Ministry of Education), Peking UniversityCancer Hospital & Institute, Beijing, China

2 Beijing Novogene Bioinformatics Technology Co., Ltd,Beijing, China

3 Department of Gastrointestinal Surgery, Key laboratory ofCarcinogenesis and Translational Research (Ministry of Education),Peking University Cancer Hospital & Institute, Beijing, China

https://doi.org/10.1007/s12253-020-00844-w

/ Published online: 3 July 2020

Pathology & Oncology Research (2020) 26:2577–2585

http://crossmark.crossref.org/dialog/?doi=10.1007/s12253-020-00844-w&domain=pdfhttps://doi.org/10.1007/s12253-020-00844-wmailto:ji_jiafu@hotmail.commailto:Dongm_lin@126.com

Thus, an assay and algorithm shall need to be fully vali-dated before being clinically used.

Human epidermal receptor growth factor (HER2) amplifi-cation or an overexpression occurs in approximately 18–20%of breast cancers, and in nearly 20% of gastric or gastroesoph-ageal junction (GEJ) cancers [8, 9]. Several methods havebeen recommended for HER2 amplification assessment, in-cluding in situ hybridization (ISH) techniques, which evaluateHER2 status by measuring the number of HER2 gene copies,or IHC, which quantifies protein expression [9]. The ASCO/CAP has provided detailed guidelines for conducting andinterpreting HER2 status in a clinical practice. These scoringmethods classify cases into “positive”, “negative”, and“equivocal” categories [10]. According to these guidelines,equivocal HER2 status necessitates additional testing, thusincreasing the cost of patient management, and delaying thedecision to recommend HER2-targeted therapy. The correla-tion between a copy number called by NGS with an averageHER2 copy number, HER2/CEP17 ratio, or IHC score is notwell established. With increasing NGS use in a clinical prac-tice it is increasingly important to validate the amplicon-baseddetection method against the standard methodologies.

In the current study, we performed and evaluated anamplicon-based NGS assay to assess HER2 amplification inbreast and gastric cancers, by using a custom designed paneland bioinformatics pipeline. We evaluated accuracy and con-cordance of NGS detection compared with the gold-standardFISH/IHC analysis methodologies.

Materials and Methods

Study and Panel Design

For CNV detection, we designed an amplicon-based panelcovering 50 genes, which included 13 CNV genes and 6 base-line genes (supplement Tables 1, 2). Briefly, we used cell lineswith known amplifications for validating accuracy and hadgreat precision for copy number detection. Then, it was ex-panded to the FFPE samples from both the breast and gastriccancer patients. We further compared the copy number detect-ed by NGS with the FISH/IHC results from the same sample,and determined a cut-off value of NGS to determine HER2status. The study schema is summarized in the supplementaryFig. 1.

Cell Line CNV Detection

To validate CNV detection, we pooled four cell lines, eachbearing single focal gene amplification (HER2, MET, EGFRand FGFR3) with the matched normal cell lines (GM18511)in several dilution series (40%, 30%, 25%, 20%, 15%, 10%,5%, 4.5%, 3%). The standard materials list is shown in

supplementary Table 3. For orthogonal support, the copynumber of the molecular standard materials were also mea-sured by digital PCR using the QuantStudio 3D digital PCRsystem (Life Technology, CA, USA).

Patients and Samples

To verify our custom designed 50 gene panel in clinical ap-plication, we used FFPE samples from 280 invasive breastcancer patients and 50 gastric cancer patients obtained fromPeking University Cancer Hospital. Five FFPE slides, each5-μm thick, were obtained from breast cancer patient samplesalong with ten FFPE slides from gastric cancer. Tumors with ahigh degree of necrosis and < 1000 tumor cells were excluded.More than 80% of the samples we finally selected were sam-ples with tumor purity greater than 20%. This study was ap-proved by theMedical Ethics Committee of PekingUniversityCancer Hospital, and the investigation was performed in ac-cordance with the Declaration of Helsinki Principles. All pa-tients had signed informed consent for the tissue research, andall of clinical data and samples were deidentified prior to anal-ysis. All the experiments were carried out in accordance withthe guideline released by the National Health and FamilyPlanning Commission of the PRC.

DNA Extraction

Genomic DNA was extracted from unstained FFPE samplesusing TIANamp FFPE DNA Kit (TIANGEN, Beijing,China), according to manufacturer’s instructions. DNA wasquantified using the Qubit dsDNA HS Assay Kit (LifeTechnology, CA, USA) and the Qubit 2.0 Flurometer (LifeTechnology, CA, USA) according to recommended protocols.Quality checks were performed by testing 5 ng DNA in 1%agarose gel electrophoresis. Samples in which the main DNAstrip in agarose gel electrophoresis less than 600 bp wereexcluded. The DNA was stored at −20 °C.

NGS Library Preparation

Sequence libraries were prepared by using library preparationreagents from Life Technology, CA, USA. The amount ofDNA input was 15 ng. Libraries were constructed using acustom designed panel (50 hotspot genes). Then, theamplicons by Ion Ampliseq Library Kit 2.0 were barcodedduring library generation using the Ion Xpress BarcodeAdapters 1–96 Kit. The libraries were purified by AMPureXP beads, quantified using the Ion Library Quantitation Kit,and qualified using Agilent Bioanalyzer 2100. Then the librar-ies were pooled for sequencing. Multiplex barcoded librarieswere enriched by clonal amplification using emulsion PCR onIon Sphere particles (Ion PI™ Template OT2 200 Kit v3, LifeTechnology, CA, USA) and loaded on an Ion PI™ Chip.

2578 D. Niu et al.

Massively parallel sequencing was carried out on Ion Protonplatform using the Ion PI™ Sequencing 200 Kit v3 accordingto manufacturer’s instructions.

Sequencing and Data Analysis

Torrent Suite Software (version 4.4.3) was used to performsignal processing, base calling, quality score assignment, andadapter trimming after the sequencing reaction. High qualityreads were aligned to human genome 19 reference bytmap4.2.18 software. Quality control and coverage analysiswas performed by an in-house analysis pipeline.

Base Substitution, Short Insertion and DeletionAnalysis

TVC (Torrent Variant Caller, version 4.4) was used to callSNV and InDel variants. TVC modules use freebayes to dis-cover candidate variants combined with the hotspots file fordetecting genemutations. Somatic mutations were determinedusing the following filters: (i) the minimum coverage depthwas 100 for SNP and 200 for InDel; (ii) theminimum cutoff ofMAF was 0.01 for hotspot variants and 0.05 for others; (iii)detected SNVs and InDels also required at least 25 variant-contained reads to be reported as positive. Those combinedminimum coverage, MAF and variant-contained reads to en-sure the accuracy of variant calls.

Copy Number Variation Detection

We used an exome-like approach, rather than the averagecoverage of exon pull-down regions with read counts peramplicon, for identifying CNV. The coverage of eachamplicon was calculated as the number of reads which cov-ered more than one amplicon but mostly aligned to theamplicon. Then, the amplicon-level coverage was divided bythe median coverage of each amplicon to normalize or mini-mize inter-sample variation. The normalized amplicon-levelcoverage was also corrected by GC content to remove thedependency of coverage across the different GC profiles.Amplicons with a coverage of less than 100 × were excludedfrom analysis. The copy number ratio of each amplicon wascalculated by dividing the GC corrected amplicon-level cov-erage of tumor samples with that of the matched normal sam-ple or normal pool. In this study, we used a normal poolderived from 14 normal breast cancer patient FFPE samplesinstead of the matched normal sample for a reference of dip-loid genome comparison. The copy number ratio was thenlog-transformed to yield the log2 copy number ratio, whichwas subsequently used to determine gene amplification status.The gene level fold change was determined as the weightedaverage of amplicon-level log-copy number ratios, for whichthe weight of each amplicon was proportional to the number

of reads; basically, the reads in the matched normal samples orthe normal pool. The final copy number of gene was equal totwice the gene level fold change.

HER2 IHC and FISH Testing

HER2 amplification was determined by IHC and the dualprobe FISH test. FISH results were reported as averageHER2 copy number and HER2/CEP17 ratio. All FFPE sam-ples were reviewed by two individual pathologists to deter-mine HER2 status.

Statistics

The accuracy (sensitivity and specificity) and precision (re-peatability and reproducibility) of NGS was evaluated withstandard material result. The correlations of the copy numbercalled by NGS and that determined by digital PCR were stud-ied by using R software. Comparisons of copy numbers whichwere detected in the three runs, were analyzed using theANOVA test.

Results

Assay Performance of NGS Calls in Standard Materials

The NGS assay performance for detecting CNVwas analyzedby detecting standard materials. The copy number of standardmaterials is summarized in Supplementary Table 3. The pre-cision was assessed in inter-assay and intra-assay studies. Wefirst simultaneously ran the two libraries which had been pre-pared by two different operators (Lib1 and Lib3). Then, Lib1was done on another run (Lib2), yielding a total of 3 replicatesfor each sample. Then, it was repeatability evaluated on a per-gene basis among the replicates. Each gene had also a similarcopy number estimation in the replicated libraries (Fig. 1). Nostatistical differences in the copy number could be detectedamong the three runs (F-value = 0.022, P value = 0.9783).The coefficient of variation (CV%) for the variation in copynumber was

tumor purity than those with a low CNV (Fig. 2b). Subsequentanalyses were focused specifically on HER2 detection.

Distribution of HER2 Amplification Status in BreastCancer

All the 280 FFPE samples from breast cancer patients weretested using FISH and IHC for establishing the HER2 ampli-fication status. The IHC results showed that 31.4% (88/280)

scored as IHC 3+, 185 as IHC 2+ and 7 as IHC 1+. Thespecimens were then tested by FISH, except for one IHC 2+sample and one IHC 3+ sample, for which testing had failed.FISH results were 100% consistent with IHC for 1+ and 3+samples. For 184 IHC 2+ samples, a total of 121 negativesamples, 63 positive samples. Two pathologists reviewedthese histologic findings, and ranked 143 samples as negativeand 41 samples as positive. Overall, this led to 129 positivesamples, 151 negative samples.

Fig. 2 The correlation analysis of CNV detection by preparing libraries. aCurve regression analysis between copy number detected by NGS anddigital PCR. The copy number detected by digital PCR was used asexpected copy number. The copy number detected by NGS was used asdetected copy number. b The effect of tumor purity on the detection limit

for gene amplifications. Four cell lines each bearing single focal geneamplification (HER2, MET, EGFR and FGFR3) were pooled with theirmatched normal cell lines in several dilution series. Detection limits wereevaluated with dilution series

Fig. 1 Repeatability of copynumber variation detection. NGS-based CNV detections were ana-lyzed using standard materialsand replicated libraries

2580 D. Niu et al.

Determination and Evaluation of Cut-off toCategorize the HER2 Status in Breast Cancer UsingNGS

After the validation in the cell lines, we implemented CNVdetection in large scale clinical FFPE specimens. The NGS-detected copy number was correlated with the FISH results(Fig. 3). To assess the sensitivity for detecting CNV whencompared to the gold-standard method, we evaluated thequantitative correlation between NGS and the average HER2copy number or HER2/CEP17 ratio detected by FISH. Weused a total of 255 samples to fit a linear regression modelof NGS copy number and the FISH testing results(Supplementary Table 5), excluding those with FISH testingfailure (n = 3),HER2 status reviewed by pathologists (n = 22).The log10 copy number detected by NGS correlated witheither the log10 average HER2 copy number (signals/cell)(y = 0.044 + 0.73x, r = 0.844; add p < 0.001), and log10

HER2/CEP17 ratio (y = 0.26 + 0.74x, r = 0.815; addp < 0.001). As NCCN guideline considered average copynumber of HER2 ≥ 6.0 signals/cell as positive, if the averagecopy number is 6 then the NGS copy number is 4.09 accord-ing to this equation. Similarly, if the average copy number is4, the NGS copy number will be 3.04, and if theHER2/CEP17ratio is set as 2, the NGS copy number will be 3.03. No false-positive or false-negative samples were found when the copynumber detected by NGS was >4.09. This value was henceidentified as the confident positive cut-off.

To determine the negative cut-off value, we used 151HER2 negative samples to represent the copy number distri-bution in HER2 negative samples. The copy number datafollowed a normal distribution (P value of Shapiro-Wilk test =0.0539). We then used the mean + 3 ×MAD (mean absolutedeviation) as the negative cut-off value (Supplementary Fig.2). Two false-positive sample was found when the copy num-ber ranged between 2.91 and 4.09, and five false-negative

Fig. 3 Amplicon-based NGS to detect CNV from breast cancerspecimens identified by IHC and FISH. Representative microscopicalresults of HER2 amplification (positive, negative and equivocal HER2status) were shown by IHC (a, d, g) and dual color FISH (b, e, h). FFPE

resection specimens with identified by FISH were further analyzed byNGS (c, f, i). The y-axis shows log2 copy number ratios of each ampliconfrom each gene

2581Evaluation of Next Generation Sequencing for Detecting HER2 Copy Number in Breast and Gastric Cancers

samples were observed when the copy number was 4.09 and < 2.91, and thecopy number between 2.91 and 4.09 as weak positive status.Overall, we achieved a sensitivity of 95.35% (123/129) and aspecificity of 98.67% (149/151) compared with HER2 status,which was determined by two pathologists who consideredboth IHC and FISH testing results in breast cancer (Table 1).

Classification of HER2 Equivocal Status Samples inBreast Cancer Using NGS

Compared with the HER2 status by IHC score, all 7 IHC 1+samples were classified as HER2-negative by NGS. Among88 IHC 3+ samples, one (1/88) was identified to be negativeby NGS (copy number = 2.71). For 185 IHC 2+ samples,NGS had an 87.80% (36/41) sensitivity and a 98.61% (142/144) specificity compared with HER2 status determined bytwo pathologists after considering both IHC and FISH testing.

Seven discrepancies were found, six of which had occurredin the context of a lowered tumor content and HER2 hetero-geneity (Supplementary Fig. 3). The average HER2 copynumber of these samples was 10.6, 5.9, 5.1, 4.1, 13.48 and7.5, respectively. The HER2/CEP17 ratio was 3.2, 2.9, 4.7,2.28, 6.1 and 4.4, respectively. However, the copy number ofthese samples were < 2.91. All the six specimens displayedhigh tumor heterogeneity, which was thought to influenceNGS accuracy.

CNV Detection in Gastric Cancer Using NGS

A total number of 50 gastric cancer patients were enrolled inthis study. In the initial 39 fully extracted FFPE samples, NGSidentified 11 out of 17 HER2 positive samples, and all the 22HER2 negative samples. Although the positive predictive val-ue (PPV) was 100%, NGS categorized all HER2 negativesamples correctly, with a low sensitivity (64.71%, 11/17).Compared with breast cancer, gastric cancer appeared to bemore heterogeneous. We hence decided to evaluate whether amacrodissection based IHC result would help to increase thesensitivity. There were 28 HER2 amplified samples

sequenced after macrodissection, and sensitivity was 57.14%(16/28). Although the NGS-detected copy number of mostsamples increased after macrodissection (Fig. 4), it was notsufficient to improve the accuracy of the NGS assay forHER2detection in gastric cancer (Table 2).

Concurrent Detection of Other Somatic Alterations inthe Clinical FFPE Samples

Overall, 118 breast cancer samples with at least one genemutation detected were include in this study. The most com-mon mutations were PIK3CA (109/280), TP53 (4/280), AKT1(3/280), KRAS (3/280), HER2 (2/280), ALK (1/280), EGFR(1/280), and RET (1/280). Among these mutations, annotatedbased on the OncoKb Knowledge Base [11], PIK3CA, AKT1and HER2 mutations were labelled with a 3A level(Compelling clinical evidence supports the biomarker as be-ing predictive of response to a drug in this indication, butneither biomarker nor drug are standard care).When assigningsamples to the level of the most actionable alteration, 40.71%(114/280) patients harbored at least one potentially actionablealteration, which may be a response to a drug, although this isnot been defined as standard care so far.

In gastric cancer, MSH3 was the most common mutation,occurring in 74% (37/50) patients. The frequency of this mu-tation was higher in gastric cancer without HER2 amplifica-tions (HER2-negative patients: 91%, HER2-positive patients:61%; p = 0.036, Chi-square test).

Discussion

HER2 amplification has both predictive and prognostic valuefor breast cancer. Currently, it is regarded as the only biomark-er established for selecting specific therapy for patients withadvanced gastric cancer [12, 13]. The current gold-standardapproach for assessing HER2 amplification status is based onthe IHC and ISH techniques. However, debate continues overthe best way to relate HER2 test results with treatment out-comes. One drawback of the HER2 test is that the scoringsystem used to determine HER2 status is subjective. In thisstudy, we developed an amplicon-based NGS panel to accu-rately detect clinically relevant copy number alterations.EvaluatingHER2 copy number with NGS in our study yieldedcomparable results to the gold-standard FISH/IHC analyses inbreast cancer patients, achieving 95.35% sensitivity and98.67% specificity.

The cell line dilution study showed that the detection foramplification is strongly influenced by tumor purity. Asshown in Fig. 2b, copy number was reduced in parallel withdecrease of tumor purity. For example, the initial copy numberwas about 4 for FGFR1 at 80% tumor purity, whilst the copynumber was approximately 2.45 at 20% tumor purity. Thus, if

Table 1 Performance of NGS assay inHER2 amplification detection inbreast cancer patients

Platform FISH Total Sensitivity Specificity Concordance

+ –

NGS + 123 2 125 95.35% 98.67% 97.14%

– 6 149 155

Total 129 151 280

2582 D. Niu et al.

the initial amplification level was low, it would not be detect-able at a low tumor purity. Other studies also shown thatsamples with poor quality or lowDNA content can yield noisyCN plots, thus limiting accurate assessment. The performancewas also affected with lower CNVs (6–7 copies) and in sam-ples with poor purity (20–30%) [14]. Thus, determining ade-quate tumor purity would be mandatory, for an accurate as-sessment. Furthermore, an alternate method shall be recom-mended for cases with amplification in combination with lowtumor content.

In addition to providing an accurate copy number, we alsoneed to transfer the continuous copy number value into abinary amplification status. In this study, we used known neg-ative samples to determine the negative cut-off value of NGSand established that 2.91 yielded a 95.35% sensitivity and98.67% specificity when compared to the gold-standard meth-od. To be more confident with NGS CNV detection, we usedthe established cut-off value for FISH, to estimate a positivecut-off value of NGS according to their correlation. It showedthat a copy number using NGS higher than 4.09 corresponded

Fig. 4 Effect of macrodissection on HER2 detection by NGS in gastriccancer. a Diagram showing macrodissection of FFPE tissue area of HEstained section of gastric cancer. b Serial section from the same specimen

for HER2 IHC. cHER2 FISH (red signal,HER2; green signal,CEP17). dCNV detection by NGS after macrodissection. e CNV detection by NGSbefore macrodissection

Table 2 Performance of NGSassay in HER2 amplificationdetection in gastric cancer

Platform FISH Total Sen Spe Con PPV NPV

+ –

NGS Macro dissection + 16 0 16 57.14% – 57.14% 100% 0.00%– 12 0 12

Full extraction + 11 0 11 64.71% 100% 84.62% 100% 78.57%– 6 22 28

2583Evaluation of Next Generation Sequencing for Detecting HER2 Copy Number in Breast and Gastric Cancers

to an average HER2 copy number higher than 6 signs/cell.Thus, we recommended an NGS copy number between 2.91and 4.09 to take an additional reflex test. However, it could beforeseen that larger validated samples would reduce thisgreyscale. In fact, our methods derive a nearly identical cutoffto call HER2 amplificated by NGS as the large commercialNGS provided Foundation Medicine, Inc. which uses 4 NGS-derived copies to call amplification in HER2.

According to updated 2018 ASCO/CAP guideline, con-comitant IHC assays are required to arrive at the most accurateHER2 status designation after HER2 FISH equivocal results.Currently, our research gave preliminary suggestions, whetherdual-probe ISH group 2 to 4 in 2018 ASCO/CAP guidelinecan be considered for inclusion in the negative. NGS mightprovide accurate assessment for the HER2 status designation,and thus reduces the risk of misdiagnosis, and further verifi-cation is required.

NGS still has some shortcomings for detection mutations,e.g., CNV detection accuracy was based on accuratelyassessing coverage depth of genes, which can be biased byhigh GC content and repetitive regions. A previous study re-ported that the number of amplicons per gene on the panelmay influence performance of CNV detection [7]. The highernumber of amplicons per gene would have somehow de-creased the variance in CNV assessment, but it would havealso restricted the list of assessable genes.

In some cases, metastatic tumors have different molecularalterations from the primary tumors. Even in breast cancer, adiscrepancy of HER2 status between primary tumor and dis-tant metastases has been observed in 7–26% of patients [15].Regarding gastric cancer, tumor heterogeneity could be pre-cisely identified using ctDNA [16] or planning reflex testingon residual materials or additional tumor blocks. Similar find-ings have been reported in advanced gastric cancer patients,whose primary tumors were found to be HER2 negative, butwhose circulating tumor cells displayed HER2 amplification[17].

Conclusions

Our study demonstrate that an optimized NGS-based test canaccurately detect most clinically targetable CNV in a broadspectrum of cancer patients. NGS-based HER2 assessmentmay decrease the equivocal HER2 determinations in breastcancer patients assessed by FISH/IHC. However, due to het-erogeneity of gastric cancer tumor tissue, detection of HER2amplification by NGS seems still problematic in thismalignancy.

Acknowledgments We acknowledge all other studies that support ourwork and were not cited due to length limitations.We would like to thankall the patients who have participated in this study.We are also thankful to

the doctors from departments of oncology, doctors from departments ofpathology for technics and staff of the Novogene Bioinformatics. Thisarticle has been edited by Jeremy Dean Chapnick, English LanguageEditor, AME Publishing Company.

Availability of Data and Materials The datasets supporting the conclu-sions of this article are included within the article and additional files.

Authors’ Contributions Jiafu Ji, Dongmei Lin and Yang Yu conceivedthe research idea and designed this study. Zhongwu Li, Lixin Zhou andLing Jia performed sample collection. Yun Gao, Donfeng Niu, Lei Li,Wanchun Zang and Lianju Gao analyzed and interpreted data. All authorsparticipated in writing, review and final approval of the manuscript.

Funding This study was funded by National Natural Science Foundationof China (81472743).

Ethics Approval and Consent to Participate

This study was approved by the ethics committee of Peking UniversityCancer Hospital (2014TW04), and written informed consent was obtain-ed from all patients.

Competing Interests Lei Li, Wanchun Zang, Lianju Gao, GuanhuaRao, Yun Gao, Gang Cheng and Yang Yu were employee ofNovogene Bioinformatics Technology Co., Ltd., Beijing, China. No po-tential conflicts of interest were disclosed by the other authors. SJK hasreceived consulting/advisory fees from Lilly Oncology, Astellas, BostonBiomedical and Foundation Medicine, Inc. SJK has stock/equity in TPTherapeutics and receives institutional research funding from Merck,Leap therapeutics, and Astellas Pharmaceuticals.

Abbreviations NGS, next generation sequencing; CNV, copy numbervariation; IHC, immunohistochemistry; FISH, fluorescence in situ hy-bridization; HER2, human epidermal receptor growth factor; FFPE, for-malin-fixed paraffin-embedded

Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, aslong as you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons licence, and indicate ifchanges weremade. The images or other third party material in this articleare included in the article's Creative Commons licence, unless indicatedotherwise in a credit line to the material. If material is not included in thearticle's Creative Commons licence and your intended use is notpermitted by statutory regulation or exceeds the permitted use, you willneed to obtain permission directly from the copyright holder. To view acopy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

References

1. Hirsch FR, Scagliotti GV, Mulshine JL, Kwon R, Curran WJ Jr,Wu YL, Paz-Ares L (2017) Lung cancer: current therapies and newtargeted treatments. Lancet 389(10066):299–311. https://doi.org/10.1016/S0140-6736(16)30958-8

2. Rehm HL, Bale SJ, Bayrak-Toydemir P, Berg JS, Brown KK,Deignan JL, Friez MJ, Funke BH, Hegde MR, Lyon E, WorkingGroup of the American College of Medical G, GenomicsLaboratory Quality Assurance C (2013) ACMG clinical laboratory

2584 D. Niu et al.

http://creativecommons.org/licenses/by/4.0/https://doi.org/10.1016/S0140-6736(16)30958-8https://doi.org/10.1016/S0140-6736(16)30958-8

standards for next-generation sequencing. Genet Med 15(9):733–747. https://doi.org/10.1038/gim.2013.92

3. Hamblin A, Wordsworth S, Fermont JM, Page S, Kaur K, CampsC, Kaisaki P, Gupta A, Talbot D,MiddletonM, Henderson S, CuttsA, Vavoulis DV, Housby N, Tomlinson I, Taylor JC, Schuh A(2017) Clinical applicability and cost of a 46-gene panel for geno-mic analysis of solid tumours: retrospective validation and prospec-tive audit in the UK National Health Service. PLoS Med 14(2):e1002230. https://doi.org/10.1371/journal.pmed.1002230

4. Pfarr N, Penzel R, Endris V, Lier C, Flechtenmacher C, VolckmarAL, Kirchner M, Budczies J, Leichsenring J, Herpel E, Noske A,Weichert W, Schneeweiss A, Schirmacher P, Sinn HP, StenzingerA (2017) Targeted next-generation sequencing enables reliable de-tection of HER2 (ERBB2) status in breast cancer and providesancillary information of clinical relevance. Gene ChromosomeCancer 56(4):255–265. https://doi.org/10.1002/gcc.22431

5. Yamamoto T, Shimojima K, Ondo Y, Imai K, Chong PF, Kira R,Amemiya M, Saito A, Okamoto N (2016) Challenges in detectinggenomic copy number aberrations using next-generation sequenc-ing data and the eXome hidden Markov model: a clinical exome-first diagnostic approach. Hum Genome Var 3:16025. https://doi.org/10.1038/hgv.2016.25

6. Ellingford JM, Campbell C, Barton S, Bhaskar S, Gupta S, TaylorRL, Sergouniotis PI, Horn B, Lamb JA, Michaelides M, WebsterAR, Newman WG, Panda B, Ramsden SC, Black GC (2017)Validation of copy number variation analysis for next-generationsequencing diagnostics. Eur J Hum Genet 25(6):719–724. https://doi.org/10.1038/ejhg.2017.42

7. Grasso C, Butler T, Rhodes K, Quist M, Neff TL, Moore S,Tomlins SA, Reinig E, Beadling C, Andersen M, Corless CL(2015) Assessing copy number alterations in targeted, amplicon-based next-generation sequencing data. J Mol Diagn 17(1):53–63.https://doi.org/10.1016/j.jmoldx.2014.09.008

8. Van Cutsem E, Bang YJ, Feng-Yi F, Xu JM, Lee KW, Jiao SC,Chong JL, Lopez-Sanchez RI, Price T, Gladkov O, Stoss O, Hill J,Ng V, Lehle M, Thomas M, Kiermaier A, Ruschoff J (2015) HER2screening data from ToGA: targeting HER2 in gastric and gastro-esophageal junction cancer. Gastric Cancer 18(3):476–484. https://doi.org/10.1007/s10120-014-0402-y

9. Owens MA, Horten BC, Da Silva MM (2004) HER2 amplificationratios by fluorescence in situ hybridization and correlation withimmunohistochemistry in a cohort of 6556 breast cancer tissues.Clin Breast Cancer 5(1):63–69. https://doi.org/10.3816/cbc.2004.n.011

10. Wolff AC, Hammond MEH, Allison KH, Harvey BE, Mangu PB,Bartlett JMS, Bilous M, Ellis IO, Fitzgibbons P, Hanna W, JenkinsRB, Press MF, Spears PA, Vance GH, Viale G, McShane LM,Dowsett M (2018) Human epidermal growth factor receptor 2 test-ing in breast cancer: American Society of Clinical Oncology/College of American Pathologists Clinical Practice GuidelineFocused Update. J Clin Oncol 36(20):2105–2122. https://doi.org/10.1200/JCO.2018.77.8738

11. Chakravarty D, Gao J, Phillips SM, Kundra R, Zhang H, Wang J,Rudolph JE, Yaeger R, Soumerai T, Nissan MH, Chang MT,Chandarlapaty S, Traina TA, Paik PK, Ho AL, Hantash FM,Grupe A, Baxi SS, Callahan MK, Snyder A, Chi P, Danila D,Gounder M, Harding JJ, Hellmann MD, Iyer G, Janjigian Y,

Kaley T, Levine DA, Lowery M, Omuro A, Postow MA,Rathkopf D, Shoushtari AN, Shukla N, Voss M, Paraiso E, ZehirA, Berger MF, Taylor BS, Saltz LB, Riely GJ, Ladanyi M, HymanDM, Baselga J, Sabbatini P, Solit DB, Schultz N (2017) OncoKB: aprecision oncology Knowledge Base. JCO precision oncology2017. https://doi.org/10.1200/PO.17.00011

12. Bartley AN, Washington MK, Colasacco C, Ventura CB, IsmailaN, Benson AB 3rd, Carrato A, Gulley ML, Jain D, Kakar S,Mackay HJ, Streutker C, Tang L, Troxell M, Ajani JA (2017)HER2 Testing and Clinical Decision Making in GastroesophagealAdenocarcinoma: Guideline From the College of AmericanPathologists, American Society for Clinical Pathology, and theAmerican Society of Clinical Oncology. J Clin Oncol 35(4):446–464. https://doi.org/10.1200/JCO.2016.69.4836

13. Wolff AC, Hammond ME, Hicks DG, Dowsett M, McShane LM,Allison KH, Allred DC, Bartlett JM, Bilous M, Fitzgibbons P,Hanna W, Jenkins RB, Mangu PB, Paik S, Perez EA, Press MF,Spears PA, Vance GH, Viale G, Hayes DF, American Society ofClinical O, College of American P (2014) Recommendations forhuman epidermal growth factor receptor 2 testing in breast cancer:American Society of Clinical Oncology/College of AmericanPathologists clinical practice guideline update. Arch Pathol LabMed 138(2):241–256. https://doi.org/10.5858/arpa.2013-0953-SA

14. Frampton GM, Fichtenholtz A, Otto GA, Wang K, Downing SR,He J, Schnall-Levin M, White J, Sanford EM, An P, Sun J, Juhn F,Brennan K, Iwanik K, Maillet A, Buell J, White E, Zhao M,Balasubramanian S, Terzic S, Richards T, Banning V, Garcia L,Mahoney K, Zwirko Z, Donahue A, Beltran H, Mosquera JM,Rubin MA, Dogan S, Hedvat CV, Berger MF, Pusztai L, LechnerM, Boshoff C, Jarosz M, Vietz C, Parker A, Miller VA, Ross JS,Curran J, Cronin MT, Stephens PJ, Lipson D, Yelensky R (2013)Development and validation of a clinical cancer genomic profilingtest based on massively parallel DNA sequencing. Nat Biotechnol31(11):1023–1031. https://doi.org/10.1038/nbt.2696

15. Fehm T, Muller V, Aktas B, Janni W, Schneeweiss A, Stickeler E,Lattrich C, Lohberg CR, Solomayer E, Rack B, Riethdorf S, KleinC, Schindlbeck C, Brocker K, Kasimir-Bauer S, Wallwiener D,Pantel K (2010) HER2 status of circulating tumor cells in patientswithmetastatic breast cancer: a prospective, multicenter trial. BreastCancer Res Treat 124(2):403–412. https://doi.org/10.1007/s10549-010-1163-x

16. Gao J, Wang H, Zang W, Li B, Rao G, Li L, Yu Y, Li Z, Dong B,Lu Z, Jiang Z, Shen L (2017) Circulating tumor DNA functions asan alternative for tissue to overcome tumor heterogeneity in ad-vanced gastric cancer. Cancer Sci 108(9):1881–1887. https://doi.org/10.1111/cas.13314

17. Mishima Y, Matsusaka S, Chin K, Mikuniya M, Minowa S,Takayama T, Shibata H, Kuniyoshi R, Ogura M, Terui Y,Mizunuma N, Hatake K (2017) Detection of HER2 amplificationin circulating tumor cells of HER2-negative gastric Cancer patients.Target Oncol 12(3):341–351. https://doi.org/10.1007/s11523-017-0493-6

Publisher’s Note Springer Nature remains neutral with regard to juris-dictional claims in published maps and institutional affiliations.

2585Evaluation of Next Generation Sequencing for Detecting HER2 Copy Number in Breast and Gastric Cancers

https://doi.org/10.1038/gim.2013.92https://doi.org/10.1371/journal.pmed.1002230https://doi.org/10.1002/gcc.22431https://doi.org/10.1038/hgv.2016.25https://doi.org/10.1038/hgv.2016.25https://doi.org/10.1038/ejhg.2017.42https://doi.org/10.1038/ejhg.2017.42https://doi.org/10.1016/j.jmoldx.2014.09.008http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/https://doi.org/10.3816/cbc.2004.n.011https://doi.org/10.3816/cbc.2004.n.011https://doi.org/10.1200/JCO.2018.77.8738https://doi.org/10.1200/JCO.2018.77.8738https://doi.org/10.1200/PO.17.00011https://doi.org/10.1200/JCO.2016.69.4836https://doi.org/10.5858/arpa.2013-0953-SAhttps://doi.org/10.1038/nbt.2696https://doi.org/10.1007/s10549-010-1163-xhttps://doi.org/10.1007/s10549-010-1163-xhttps://doi.org/10.1111/cas.13314https://doi.org/10.1111/cas.13314https://doi.org/10.1007/s11523-017-0493-6https://doi.org/10.1007/s11523-017-0493-6

Evaluation of Next Generation Sequencing for Detecting HER2 Copy Number in Breast and Gastric CancersAbstractBackgroundMaterials and MethodsStudy and Panel DesignCell Line CNV DetectionPatients and SamplesDNA ExtractionNGS Library PreparationSequencing and Data AnalysisBase Substitution, Short Insertion and Deletion AnalysisCopy Number Variation DetectionHER2 IHC and FISH TestingStatistics

ResultsAssay Performance of NGS Calls in Standard MaterialsDistribution of HER2 Amplification Status in Breast CancerDetermination and Evaluation of Cut-off to Categorize the HER2 Status in Breast Cancer Using NGSClassification of HER2 Equivocal Status Samples in Breast Cancer Using NGSCNV Detection in Gastric Cancer Using NGSConcurrent Detection of Other Somatic Alterations in the Clinical FFPE Samples

DiscussionConclusionsReferences